Machine Learning\nReveals Signatures of Promiscuous\nMicrobial Amidases for Micropollutant Biotransformations

Organic micropollutants, including pharmaceuticals, personal\ncare\nproducts, pesticides, and food additives, are widespread in the environment,\ncausing potentially toxic effects. Human waste is a direct source\nof micropollutants, with the majority of pharmaceuticals being excreted\nthrough urine. Urine contains its own microbiota with the potential\nto catalyze micropollutant biotransformations. Amidase signature (AS)\nenzymes are known for their promiscuous activity in micropollutant\nbiotransformations, but the potential for AS enzymes from the urinary\nmicrobiota to transform micropollutants is not known. Moreover, the\ncharacterization of AS enzymes to identify key chemical and enzymatic\nfeatures associated with biotransformation profiles is critical for\ndeveloping benign-by-design chemicals and micropollutant removal strategies.\nHere, to uncover the signatures of AS enzyme–substrate specificity,\nwe tested 17 structurally diverse compounds against a targeted enzyme\nlibrary consisting of 40 AS enzyme homologues from diverse urine microbial\nisolates. The most promiscuous enzymes were active on nine different\nsubstrates, while 16 enzymes had activity on at least one substrate\nand exhibited diverse substrate specificities. Using an interpretable\ngradient boosting machine learning model, we identified chemical and\namino acid features associated with AS enzyme biotransformations.\nKey chemical features from our substrates included the molecular weight\nof the amide carbonyl substituent and the number of formal charges\nin the molecule. Four of the identified amino acid features were located\nin close proximity to the substrate tunnel entrance. Overall, this\nwork highlights the understudied potential of urine-derived microbial\nAS enzymes for micropollutant biotransformation and offers insights\ninto substrate and protein features associated with micropollutant\nbiotransformations for future environmental applications.